Coverage for tests/test_imageDecorrelation.py: 15%

266 statements  

« prev     ^ index     » next       coverage.py v6.5.0, created at 2022-12-01 21:04 +0000

1# 

2# LSST Data Management System 

3# Copyright 2016-2017 AURA/LSST. 

4# 

5# This product includes software developed by the 

6# LSST Project (http://www.lsst.org/). 

7# 

8# This program is free software: you can redistribute it and/or modify 

9# it under the terms of the GNU General Public License as published by 

10# the Free Software Foundation, either version 3 of the License, or 

11# (at your option) any later version. 

12# 

13# This program is distributed in the hope that it will be useful, 

14# but WITHOUT ANY WARRANTY; without even the implied warranty of 

15# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 

16# GNU General Public License for more details. 

17# 

18# You should have received a copy of the LSST License Statement and 

19# the GNU General Public License along with this program. If not, 

20# see <https://www.lsstcorp.org/LegalNotices/>. 

21import unittest 

22 

23import numpy as np 

24 

25import lsst.utils.tests 

26from lsst.utils.tests import methodParameters 

27import lsst.afw.image as afwImage 

28import lsst.afw.geom as afwGeom 

29import lsst.afw.math as afwMath 

30import lsst.geom as geom 

31import lsst.meas.algorithms as measAlg 

32import lsst.daf.base as dafBase 

33 

34from lsst.ip.diffim.imageDecorrelation import (DecorrelateALKernelTask, 

35 DecorrelateALKernelConfig, 

36 DecorrelateALKernelMapReduceConfig, 

37 DecorrelateALKernelSpatialConfig, 

38 DecorrelateALKernelSpatialTask) 

39from lsst.ip.diffim.imageMapReduce import ImageMapReduceTask 

40 

41try: 

42 type(verbose) 

43except NameError: 

44 verbose = False 

45 

46 

47def setup_module(module): 

48 lsst.utils.tests.init() 

49 

50 

51def singleGaussian2d(x, y, xc, yc, sigma_x=1., sigma_y=1., theta=0., ampl=1.): 

52 """! Generate a 2-d Gaussian, possibly elongated and rotated, on a grid of pixel 

53 coordinates given by x,y. 

54 @param x,y each a 1-d numpy.array containing x- and y- coordinates for independent variables, 

55 for example `np.arange(-16, 15)`. 

56 @param xc,yc each a float giving the centroid of the gaussian 

57 @param sigma_x,sigma_y each a float giving the sigma of the gaussian 

58 @param theta a float giving the rotation of the gaussian (degrees) 

59 @param ampl a float giving the amplitude of the gaussian 

60 @return a 2-d numpy.array containing the normalized 2-d Gaussian 

61 

62 @Note this can be done in `astropy.modeling` but for now we have it explicitly here. 

63 """ 

64 theta = (theta/180.) * np.pi 

65 cos_theta2, sin_theta2 = np.cos(theta)**2., np.sin(theta)**2. 

66 sigma_x2, sigma_y2 = sigma_x**2., sigma_y**2. 

67 a = cos_theta2/(2.*sigma_x2) + sin_theta2/(2.*sigma_y2) 

68 b = -(np.sin(2.*theta))/(4.*sigma_x2) + (np.sin(2.*theta))/(4.*sigma_y2) 

69 c = sin_theta2/(2.*sigma_x2) + cos_theta2/(2.*sigma_y2) 

70 xxc, yyc = x-xc, y-yc 

71 out = np.exp(-(a*(xxc**2.) + 2.*b*xxc*yyc + c*(yyc**2.))) 

72 out /= out.sum() 

73 return out 

74 

75 

76def makeFakeImages(size=(256, 256), svar=0.04, tvar=0.04, psf1=3.3, psf2=2.2, offset=None, 

77 psf_yvary_factor=0., varSourceChange=1/50., theta1=0., theta2=0., 

78 n_sources=500, seed=66, verbose=False): 

79 """Deprecated. Use ``test_zogy : makeFakeImages``. DM-25115: This function is kept for 

80 numerical compatibility with existing test cases only. It has x,y axes handling 

81 and PSF centering bugs. 

82 

83 Parameters 

84 ---------- 

85 size : `tuple` of `int` 

86 Image pixel size (x,y). Pixel coordinates are set to 

87 (-size[0]//2:size[0]//2, -size[1]//2:size[1]//2) 

88 svar, tvar : `float`, optional 

89 Per pixel variance of the added noise. 

90 psf1, psf2 : `float`, optional 

91 std. dev. of (Gaussian) PSFs for the two images in x,y direction. Default is 

92 [3.3, 3.3] and [2.2, 2.2] for im1 and im2 respectively. 

93 offset : `float`, optional 

94 add a constant (pixel) astrometric offset between the two images. 

95 psf_yvary_factor : `float`, optional 

96 psf_yvary_factor vary the y-width of the PSF across the x-axis of the science image (zero, 

97 the default, means no variation) 

98 varSourceChange : `float`, optional 

99 varSourceChange add this amount of fractional flux to a single source closest to 

100 the center of the science image. 

101 theta1, theta2: `float`, optional 

102 PSF Gaussian rotation angles in degrees. 

103 n_sources : `int`, optional 

104 The number of sources to add to the images. If zero, no sources are 

105 generated just background noise. 

106 seed : `int`, optional 

107 Random number generator seed. 

108 verbose : `bool`, optional 

109 Print some actual values. 

110 

111 Returns 

112 ------- 

113 im1, im2 : `lsst.afw.image.Exposure` 

114 The science and template exposures. 

115 

116 Notes 

117 ----- 

118 If ``n_sources > 0`` and ``varSourceChange > 0.`` exactly one source, 

119 that is closest to the center, will have different fluxes in the two 

120 generated images. The flux on the science image will be higher by 

121 ``varSourceChange`` fraction. 

122 

123 Having sources near the edges really messes up the 

124 fitting (probably because of the convolution). So we make sure no 

125 sources are near the edge. 

126 

127 Also it seems that having the variable source with a large 

128 flux increase also messes up the fitting (seems to lead to 

129 overfitting -- perhaps to the source itself). This might be fixed by 

130 adding more constant sources. 

131 """ 

132 np.random.seed(seed) 

133 

134 psf1 = [3.3, 3.3] if psf1 is None else psf1 

135 if not hasattr(psf1, "__len__") and not isinstance(psf1, str): 

136 psf1 = [psf1, psf1] 

137 psf2 = [2.2, 2.2] if psf2 is None else psf2 

138 if not hasattr(psf2, "__len__") and not isinstance(psf2, str): 

139 psf2 = [psf2, psf2] 

140 offset = [0., 0.] if offset is None else offset # astrometric offset (pixels) between the two images 

141 if verbose: 

142 print('Science PSF:', psf1, theta1) 

143 print('Template PSF:', psf2, theta2) 

144 print(np.sqrt(psf1[0]**2 - psf2[0]**2)) 

145 print('Offset:', offset) 

146 

147 xim = np.arange(-size[0]//2, size[0]//2, 1) 

148 yim = np.arange(-size[1]//2, size[1]//2, 1) 

149 x0im, y0im = np.meshgrid(yim, xim) 

150 im1 = np.random.normal(scale=np.sqrt(svar), size=x0im.shape) # variance of science image 

151 im2 = np.random.normal(scale=np.sqrt(tvar), size=x0im.shape) # variance of template 

152 

153 if n_sources > 0: 

154 fluxes = np.random.uniform(50, 30000, n_sources) 

155 xposns = np.random.uniform(xim.min()+16, xim.max()-5, n_sources) 

156 yposns = np.random.uniform(yim.min()+16, yim.max()-5, n_sources) 

157 

158 # Make the source closest to the center of the image the one that increases in flux 

159 ind = np.argmin(xposns**2. + yposns**2.) 

160 

161 # vary the y-width of psf across x-axis of science image (zero means no variation): 

162 psf1_yvary = psf_yvary_factor * (yim.mean() - yposns) / yim.max() 

163 if verbose: 

164 print('PSF y spatial-variation:', psf1_yvary.min(), psf1_yvary.max()) 

165 

166 for i in range(n_sources): 

167 flux = fluxes[i] 

168 tmp = flux * singleGaussian2d(x0im, y0im, xposns[i], yposns[i], psf2[0], psf2[1], theta=theta2) 

169 im2 += tmp 

170 if i == ind: 

171 flux += flux * varSourceChange 

172 tmp = flux * singleGaussian2d(x0im, y0im, xposns[i]+offset[0], yposns[i]+offset[1], 

173 psf1[0], psf1[1]+psf1_yvary[i], theta=theta1) 

174 im1 += tmp 

175 

176 im1_psf = singleGaussian2d(x0im, y0im, 0, 0, psf1[0], psf1[1], theta=theta1) 

177 im2_psf = singleGaussian2d(x0im, y0im, offset[0], offset[1], psf2[0], psf2[1], theta=theta2) 

178 

179 def makeWcs(offset=0): 

180 """ Make a fake Wcs 

181 

182 Parameters 

183 ---------- 

184 offset : float 

185 offset the Wcs by this many pixels. 

186 """ 

187 # taken from $AFW_DIR/tests/testMakeWcs.py 

188 metadata = dafBase.PropertySet() 

189 metadata.set("SIMPLE", "T") 

190 metadata.set("BITPIX", -32) 

191 metadata.set("NAXIS", 2) 

192 metadata.set("NAXIS1", 1024) 

193 metadata.set("NAXIS2", 1153) 

194 metadata.set("RADESYS", 'FK5') 

195 metadata.set("EQUINOX", 2000.) 

196 metadata.setDouble("CRVAL1", 215.604025685476) 

197 metadata.setDouble("CRVAL2", 53.1595451514076) 

198 metadata.setDouble("CRPIX1", 1109.99981456774 + offset) 

199 metadata.setDouble("CRPIX2", 560.018167811613 + offset) 

200 metadata.set("CTYPE1", 'RA---SIN') 

201 metadata.set("CTYPE2", 'DEC--SIN') 

202 metadata.setDouble("CD1_1", 5.10808596133527E-05) 

203 metadata.setDouble("CD1_2", 1.85579539217196E-07) 

204 metadata.setDouble("CD2_2", -5.10281493481982E-05) 

205 metadata.setDouble("CD2_1", -8.27440751733828E-07) 

206 return afwGeom.makeSkyWcs(metadata) 

207 

208 def makeExposure(imgArray, psfArray, imgVariance): 

209 """! Convert an image numpy.array and corresponding PSF numpy.array into an exposure. 

210 

211 Add the (constant) variance plane equal to `imgVariance`. 

212 

213 @param imgArray 2-d numpy.array containing the image 

214 @param psfArray 2-d numpy.array containing the PSF image 

215 @param imgVariance variance of input image 

216 @return a new exposure containing the image, PSF and desired variance plane 

217 """ 

218 # All this code to convert the template image array/psf array into an exposure. 

219 bbox = geom.Box2I(geom.Point2I(0, 0), geom.Point2I(imgArray.shape[1]-1, imgArray.shape[0]-1)) 

220 im1ex = afwImage.ExposureD(bbox) 

221 im1ex.getMaskedImage().getImage().getArray()[:, :] = imgArray 

222 im1ex.getMaskedImage().getVariance().getArray()[:, :] = imgVariance 

223 psfBox = geom.Box2I(geom.Point2I(-12, -12), geom.Point2I(12, 12)) # a 25x25 pixel psf 

224 psf = afwImage.ImageD(psfBox) 

225 psfBox.shift(geom.Extent2I(size[0]//2, size[1]//2)) 

226 im1_psf_sub = psfArray[psfBox.getMinX():psfBox.getMaxX()+1, psfBox.getMinY():psfBox.getMaxY()+1] 

227 psf.getArray()[:, :] = im1_psf_sub 

228 psfK = afwMath.FixedKernel(psf) 

229 psfNew = measAlg.KernelPsf(psfK) 

230 im1ex.setPsf(psfNew) 

231 wcs = makeWcs() 

232 im1ex.setWcs(wcs) 

233 return im1ex 

234 

235 im1ex = makeExposure(im1, im1_psf, svar) # Science image 

236 im2ex = makeExposure(im2, im2_psf, tvar) # Template 

237 

238 return im1ex, im2ex 

239 

240 

241def estimatePixelCorrelation(B, nDist=40, convEdge=17): 

242 """Estimate correlation as a function of pixel distance in the image 

243 by sampling pixel pairs. 

244 

245 Parameters 

246 ---------- 

247 B : `numpy.ndarray` of N x N `float` elements 

248 Noise only image with zero pixel expectation value and identical variance 

249 in all pixels. Must have equal dimensions. 

250 nDist : `int`, optional 

251 Estimated distances goes from 0 to nDist-1. 

252 nDist must be smaller than the half dimensions of B. 

253 convEdge : `int`, optional 

254 Edge width where convolution did not happen. 

255 

256 Returns 

257 ------- 

258 S : `numpy.ndarray` of nDist `float` elements 

259 Correlation from 0 to nDist-1 pix distance. Pixels are normed by their 

260 variance estimation. S[0], the autocorrelation, should be close to 1. 

261 """ 

262 S = np.zeros(nDist, dtype=float) 

263 nSample = 10000 

264 # Cannot use nDist wide edge, otherwise 2nd pixel can go off the image. 

265 # Don't bother with it. 

266 A = B/np.sqrt(np.mean(B[convEdge:-convEdge, convEdge:-convEdge] 

267 * B[convEdge:-convEdge, convEdge:-convEdge])) 

268 lEdge = nDist + convEdge 

269 rEdge = B.shape[0] - lEdge 

270 for r in range(nDist): 

271 ind1 = np.random.randint(lEdge, rEdge, (2, nSample)) 

272 ind2 = np.copy(ind1) 

273 # generate delta x,y in random directions uniformly 

274 c_dxy = np.exp(2.j*np.pi*np.random.random(nSample)) 

275 ind2[0] += np.around(np.real(c_dxy)*r).astype(int) 

276 ind2[1] += np.around(np.imag(c_dxy)*r).astype(int) 

277 S[r] = np.sum(A[ind1[0], ind1[1]] * A[ind2[0], ind2[1]])/nSample 

278 return S 

279 

280 

281class DiffimCorrectionTest(lsst.utils.tests.TestCase): 

282 """A test case for the diffim image decorrelation algorithm. 

283 """ 

284 

285 def setUp(self): 

286 self.psf1_sigma = 3.3 # sigma of psf of science image 

287 self.psf2_sigma = 2.2 # sigma of psf of template image 

288 

289 self.statsControl = afwMath.StatisticsControl() 

290 self.statsControl.setNumSigmaClip(3.) 

291 self.statsControl.setNumIter(3) 

292 self.statsControl.setAndMask(afwImage.Mask 

293 .getPlaneBitMask(["INTRP", "EDGE", "SAT", "CR", 

294 "DETECTED", "BAD", 

295 "NO_DATA", "DETECTED_NEGATIVE"])) 

296 

297 def _setUpImages(self, svar=0.04, tvar=0.04, varyPsf=0.): 

298 """Generate a fake aligned template and science image. 

299 """ 

300 

301 self.svar = svar # variance of noise in science image 

302 self.tvar = tvar # variance of noise in template image 

303 

304 self.im1ex, self.im2ex \ 

305 = makeFakeImages(svar=self.svar, tvar=self.tvar, psf1=self.psf1_sigma, psf2=self.psf2_sigma, 

306 n_sources=50, psf_yvary_factor=varyPsf, verbose=False) 

307 

308 def _setUpSourcelessImages(self, svar, tvar): 

309 """Generate noise only template and science images. 

310 """ 

311 

312 self.svar = svar # variance of noise in science image 

313 self.tvar = tvar # variance of noise in template image 

314 

315 self.im1ex, self.im2ex = makeFakeImages( 

316 svar=self.svar, tvar=self.tvar, psf1=self.psf1_sigma, psf2=self.psf2_sigma, 

317 n_sources=0, seed=22, varSourceChange=0, psf_yvary_factor=0) 

318 

319 def _computeVarianceMean(self, maskedIm): 

320 statObj = afwMath.makeStatistics(maskedIm.getVariance(), 

321 maskedIm.getMask(), afwMath.MEANCLIP, 

322 self.statsControl) 

323 mn = statObj.getValue(afwMath.MEANCLIP) 

324 return mn 

325 

326 def _computePixelVariance(self, maskedIm): 

327 statObj = afwMath.makeStatistics(maskedIm, afwMath.VARIANCECLIP, 

328 self.statsControl) 

329 var = statObj.getValue(afwMath.VARIANCECLIP) 

330 return var 

331 

332 def tearDown(self): 

333 del self.im1ex 

334 del self.im2ex 

335 

336 def _makeAndTestUncorrectedDiffim(self): 

337 """Create the (un-decorrelated) diffim, and verify that its variance is too low. 

338 """ 

339 # Create the matching kernel. We used Gaussian PSFs for im1 and im2, so we can compute the "expected" 

340 # matching kernel sigma. 

341 psf1_sig = self.im1ex.getPsf().computeShape().getDeterminantRadius() 

342 psf2_sig = self.im2ex.getPsf().computeShape().getDeterminantRadius() 

343 sig_match = np.sqrt((psf1_sig**2. - psf2_sig**2.)) 

344 # Sanity check - make sure PSFs are correct. 

345 self.assertFloatsAlmostEqual(sig_match, np.sqrt((self.psf1_sigma**2. - self.psf2_sigma**2.)), 

346 rtol=2e-5) 

347 # mKernel = measAlg.SingleGaussianPsf(31, 31, sig_match) 

348 x0 = np.arange(-16, 16, 1) 

349 y0 = x0.copy() 

350 x0im, y0im = np.meshgrid(x0, y0) 

351 matchingKernel = singleGaussian2d(x0im, y0im, -1., -1., sigma_x=sig_match, sigma_y=sig_match) 

352 kernelImg = afwImage.ImageD(matchingKernel.shape[0], matchingKernel.shape[1]) 

353 kernelImg.getArray()[:, :] = matchingKernel 

354 mKernel = afwMath.FixedKernel(kernelImg) 

355 

356 # Create the matched template by convolving the template with the matchingKernel 

357 matched_im2ex = self.im2ex.clone() 

358 convCntrl = afwMath.ConvolutionControl(False, True, 0) 

359 afwMath.convolve(matched_im2ex.getMaskedImage(), self.im2ex.getMaskedImage(), mKernel, convCntrl) 

360 

361 # Expected (ideal) variance of difference image 

362 expected_var = self.svar + self.tvar 

363 if verbose: 

364 print('EXPECTED VARIANCE:', expected_var) 

365 

366 # Create the diffim (uncorrected) 

367 # Uncorrected diffim exposure - variance plane is wrong (too low) 

368 tmp_diffExp = self.im1ex.getMaskedImage().clone() 

369 tmp_diffExp -= matched_im2ex.getMaskedImage() 

370 var = self._computeVarianceMean(tmp_diffExp) 

371 self.assertLess(var, expected_var) 

372 

373 # Uncorrected diffim exposure - variance is wrong (too low) - same as above but on pixels 

374 diffExp = self.im1ex.clone() 

375 tmp = diffExp.getMaskedImage() 

376 tmp -= matched_im2ex.getMaskedImage() 

377 var = self._computePixelVariance(diffExp.getMaskedImage()) 

378 self.assertLess(var, expected_var) 

379 

380 # Uncorrected diffim exposure - variance plane is wrong (too low) 

381 mn = self._computeVarianceMean(diffExp.getMaskedImage()) 

382 self.assertLess(mn, expected_var) 

383 if verbose: 

384 print('UNCORRECTED VARIANCE:', var, mn) 

385 

386 return diffExp, mKernel, expected_var 

387 

388 def _runDecorrelationTask(self, diffExp, mKernel, config=None): 

389 """ Run the decorrelation task on the given diffim with the given matching kernel 

390 """ 

391 task = DecorrelateALKernelTask(config=config) 

392 decorrResult = task.run(self.im1ex, self.im2ex, diffExp, mKernel) 

393 corrected_diffExp = decorrResult.correctedExposure 

394 return corrected_diffExp 

395 

396 def _testDecorrelation(self, expected_var, corrected_diffExp): 

397 """ Check that the variance of the corrected diffim matches the theoretical value. 

398 """ 

399 # Corrected diffim - variance should be close to expected. 

400 # We set the tolerance a bit higher here since the simulated images have many bright stars 

401 var = self._computePixelVariance(corrected_diffExp.getMaskedImage()) 

402 self.assertFloatsAlmostEqual(var, expected_var, rtol=0.05) 

403 

404 # Check statistics of variance plane in corrected diffim 

405 mn = self._computeVarianceMean(corrected_diffExp.getMaskedImage()) 

406 if verbose: 

407 print('CORRECTED VARIANCE:', var, mn) 

408 self.assertFloatsAlmostEqual(mn, expected_var, rtol=0.02) 

409 self.assertFloatsAlmostEqual(var, mn, rtol=0.05) 

410 return var, mn 

411 

412 def _testDiffimCorrection(self, svar, tvar, config): 

413 """ Run decorrelation and check the variance of the corrected diffim. 

414 """ 

415 self._setUpImages(svar=svar, tvar=tvar) 

416 diffExp, mKernel, expected_var = self._makeAndTestUncorrectedDiffim() 

417 corrected_diffExp = self._runDecorrelationTask(diffExp, mKernel, config) 

418 self._testDecorrelation(expected_var, corrected_diffExp) 

419 

420 @methodParameters(completeVarPlanePropagation=[False, True]) 

421 def testDiffimCorrection(self, completeVarPlanePropagation): 

422 """Test decorrelated diffim from images with different combinations of variances. 

423 """ 

424 config = DecorrelateALKernelConfig() 

425 config.completeVarPlanePropagation = completeVarPlanePropagation 

426 # Same variance 

427 self._testDiffimCorrection(svar=0.04, tvar=0.04, config=config) 

428 # Science image variance is higher than that of the template. 

429 self._testDiffimCorrection(svar=0.08, tvar=0.04, config=config) 

430 # Template variance is higher than that of the science img. 

431 self._testDiffimCorrection(svar=0.04, tvar=0.08, config=config) 

432 

433 def testNoiseDiffimCorrection(self): 

434 """Test correction by estimating correlation directly on a noise difference image. 

435 

436 Notes 

437 ------ 

438 

439 See `lsst-dm/diffimTests` notebook `DM-24371_correlation_estimate.ipynb` 

440 for further details of how the correlation looks like in the uncorrected 

441 and corrected cases and where the tolerance numbers come from. 

442 """ 

443 svar = 1. 

444 tvar = 100. 

445 # Based on DM-24371_correlation_estimate.ipynb 

446 someCorrelationThreshold = 0.2 

447 

448 self._setUpSourcelessImages(svar=svar, tvar=tvar) 

449 diffExp, mKernel, expected_var = self._makeAndTestUncorrectedDiffim() 

450 corrected_diffExp = self._runDecorrelationTask(diffExp, mKernel) 

451 

452 rho_sci = estimatePixelCorrelation(self.im1ex.getImage().getArray()) 

453 rho_rawdiff = estimatePixelCorrelation(diffExp.getImage().getArray()) 

454 rho_corrdiff = estimatePixelCorrelation(corrected_diffExp.getImage().getArray()) 

455 

456 # Autocorrelation sanity check 

457 self.assertFloatsAlmostEqual(rho_sci[0], 1., atol=0.1, rtol=None) 

458 self.assertFloatsAlmostEqual(rho_rawdiff[0], 1., atol=0.1, rtol=None) 

459 self.assertFloatsAlmostEqual(rho_corrdiff[0], 1., atol=0.1, rtol=None) 

460 

461 # Uncorrelated input check 

462 self.assertFloatsAlmostEqual(rho_sci[1:], 0., atol=0.1, rtol=None) 

463 

464 # Without correction there should be correlation up to a few pixel distance 

465 self.assertGreater(rho_rawdiff[1], someCorrelationThreshold) 

466 self.assertGreater(rho_rawdiff[2], someCorrelationThreshold) 

467 self.assertGreater(rho_rawdiff[3], someCorrelationThreshold) 

468 

469 # Uncorrelated corrected image check 

470 self.assertFloatsAlmostEqual(rho_corrdiff[1:], 0., atol=0.1, rtol=None) 

471 

472 def _runDecorrelationTaskMapReduced(self, diffExp, mKernel): 

473 """ Run decorrelation using the imageMapReducer. 

474 """ 

475 config = DecorrelateALKernelMapReduceConfig() 

476 config.borderSizeX = config.borderSizeY = 3 

477 config.reducer.reduceOperation = 'average' 

478 task = ImageMapReduceTask(config=config) 

479 decorrResult = task.run(diffExp, template=self.im2ex, science=self.im1ex, 

480 psfMatchingKernel=mKernel, forceEvenSized=True) 

481 corrected_diffExp = decorrResult.exposure 

482 return corrected_diffExp 

483 

484 def _testDiffimCorrection_mapReduced(self, svar, tvar, varyPsf=0.0): 

485 """ Run decorrelation using the imageMapReduce task, and check the variance of 

486 the corrected diffim. 

487 """ 

488 self._setUpImages(svar=svar, tvar=tvar, varyPsf=varyPsf) 

489 diffExp, mKernel, expected_var = self._makeAndTestUncorrectedDiffim() 

490 corrected_diffExp = self._runDecorrelationTaskMapReduced(diffExp, mKernel) 

491 self._testDecorrelation(expected_var, corrected_diffExp) 

492 # Also compare the diffim generated here vs. the non-ImageMapReduce one 

493 corrected_diffExp_OLD = self._runDecorrelationTask(diffExp, mKernel) 

494 self.assertMaskedImagesAlmostEqual(corrected_diffExp.getMaskedImage(), 

495 corrected_diffExp_OLD.getMaskedImage()) 

496 

497 @unittest.skip("DM-21868 ImageMapReduce usage is not yet supported") 

498 def testDiffimCorrection_mapReduced(self): 

499 """ Test decorrelated diffim when using the imageMapReduce task. 

500 Compare results with those from the original DecorrelateALKernelTask. 

501 """ 

502 # Same variance 

503 self._testDiffimCorrection_mapReduced(svar=0.04, tvar=0.04) 

504 # Science image variance is higher than that of the template. 

505 self._testDiffimCorrection_mapReduced(svar=0.04, tvar=0.08) 

506 # Template variance is higher than that of the science img. 

507 self._testDiffimCorrection_mapReduced(svar=0.08, tvar=0.04) 

508 

509 def _runDecorrelationSpatialTask(self, diffExp, mKernel, spatiallyVarying=False): 

510 """ Run decorrelation using the DecorrelateALKernelSpatialTask. 

511 """ 

512 config = DecorrelateALKernelSpatialConfig() 

513 task = DecorrelateALKernelSpatialTask(config=config) 

514 decorrResult = task.run(scienceExposure=self.im1ex, templateExposure=self.im2ex, 

515 subtractedExposure=diffExp, psfMatchingKernel=mKernel, 

516 spatiallyVarying=spatiallyVarying) 

517 corrected_diffExp = decorrResult.correctedExposure 

518 return corrected_diffExp 

519 

520 def _testDiffimCorrection_spatialTask(self, svar, tvar, varyPsf=0.0): 

521 """Run decorrelation using the DecorrelateALKernelSpatialTask, and 

522 check the variance of the corrected diffim. Do it for `spatiallyVarying` both 

523 True and False. Also compare the variances between the two `spatiallyVarying` 

524 cases. 

525 """ 

526 self._setUpImages(svar=svar, tvar=tvar, varyPsf=varyPsf) 

527 diffExp, mKernel, expected_var = self._makeAndTestUncorrectedDiffim() 

528 variances = [] 

529 for spatiallyVarying in [False, True]: 

530 corrected_diffExp = self._runDecorrelationSpatialTask(diffExp, mKernel, 

531 spatiallyVarying) 

532 var, mn = self._testDecorrelation(expected_var, corrected_diffExp) 

533 variances.append(var) 

534 self.assertFloatsAlmostEqual(variances[0], variances[1], rtol=0.03) 

535 

536 def testDiffimCorrection_spatialTask(self): 

537 """Test decorrelated diffim when using the DecorrelateALKernelSpatialTask. 

538 Compare results with those from the original DecorrelateALKernelTask. 

539 """ 

540 # Same variance 

541 self._testDiffimCorrection_spatialTask(svar=0.04, tvar=0.04) 

542 # Science image variance is higher than that of the template. 

543 self._testDiffimCorrection_spatialTask(svar=0.04, tvar=0.08) 

544 # Template variance is higher than that of the science img. 

545 self._testDiffimCorrection_spatialTask(svar=0.08, tvar=0.04) 

546 

547 

548class MemoryTester(lsst.utils.tests.MemoryTestCase): 

549 pass 

550 

551 

552if __name__ == "__main__": 552 ↛ 553line 552 didn't jump to line 553, because the condition on line 552 was never true

553 lsst.utils.tests.init() 

554 unittest.main()